Clustering is a useful approach to provide a simplified and meaningful representation of large graphs. By extracting dense communities of nodes, the "big picture" of the network organization is enlighten. Moreover, hierarchical clustering may help the user to focus on some parts of the graph which is of interest for him and that can be displayed with finer and finer details. This talk will present some strategies to analyze and display graphs based on previous node clustering. Self-organizing map like-strategies, using kernel or an modularity based optimization will be presented, as well as strategies based on a hierarchical nodes clustering. Finally, strategies trying to take the benefit of the nodes’ labels will be discussed.
		

The environmental costs of intensive farming activities are often underestimated or not traded by the market, even though they play an important role in addressing future society’s needs. The estimation of nitrogen (N) dynamics is thus an important issue which demands detailed simulation based methods and their integrated use to correctly represent complex and nonlinear interactions into cropping systems. To calculate the N2O ux and N leaching from European arable lands, a modeling framework has been developed by linking the CAPRI agro-economic dataset with the DNDC-EUROPE bio-geo-chemical model. But, despite the great power of modern calculators, their use at continental scale is often too computationally costly. By comparing several statistical methods this paper aims to design a metamodel able to approximate the expensive code of the detailed modeling approach, devising the best compromise between estimation performance and simulation speed.
		

Clustering is a useful approach to provide a simplified and meaningful representation of large graphs. By extracting dense communiites of nodes, the "big picture" of the network organizatin is enlighten. Moreover, hierarchical clustering may help the user to focus on some parts of the graph which is of interest for him and which can be displayed with finer and finer details. This talk will try to present some open issues with graph visualization based on a hierarchical nodes clustering. These issues include displaying the clusters in a coherent way between the different layers of the hierarchy or integrating information about the clustering evaluation in the visualization.
		

Des actes notariés du Moyen-Age aux réseaux sociaux actuels, l’oratrice nous montrera l’utilité des outils mathématiques que sont les graphes dans l’étude de ce type de données. Au-delà de l’aspect mathématique, c’est donc à un voyage dans le temps que l’oratrice nous convie.
		

Les graphes (ou réseaux) sont devenus des outils courants de modélisation des données relationnelles dans de nombreuses applications. Or, lorsque le nombre de sommets dépasse quelques centaines, la visualisation du graphe dans son ensemble, qui est un outil important de compréhension du réseau, est un problème complexe : les approches traditionnelles, basées sur des algorithmes de forces, s’avèrent coûteuses en temps de calcul et ne mettent pas bien en valeur la structure du réseau en parties denses (souvent appelées communautés). Dans cet exposé nous présentons plusieurs méthodes de visualisation basée sur des classifications des sommets d’un graphe : certaines combinent classification et visualisation, d’autres procèdent en deux temps mais permettent d’obtenir une visualisation hiérarchique du réseau qui autorise une exploration avec un niveau de détail progressif. Nous illustrerons ces approches sur plusieurs jeux de données publics ou bien réels.